arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2009.05314cs.PLcs.CCcs.PF
收藏

有界超线性加速的重复递归展开

Repeated Recursion Unfolding for Super-Linear Speedup within Bounds

Thom Fruehwirth

摘要:

重复递归展开是一种新的递归展开方法,它在保持所有展开规则的前提下,通过递归自身重复展开并简化递归。每一次展开都会使递归步骤的数量加倍。这将递归规则应用程序的数量减少到其对数,代价是向程序引入对数数量的展开规则。效率主要取决于展开规则内部的简化量。我们证明了一个超线性加速定理在最好的情况下,即加速超过一个常数因子。我们的优化可以降低程序的时间复杂度类。在本文中,超线性加速是有界的:它保持了一个任意的,但选择的递归步骤的数量上界。我们还报告的第一个结果与重复递归展开的原型实现。一个简单的程序转换完全消除了递归,直到选定的界限。实际的运行时改进很快达到几个数量级。

英文摘要:

Repeated recursion unfolding is a new approach that repeatedly unfolds a recursion with itself and simplifies it while keeping all unfolded rules. Each unfolding doubles the number of recursive steps covered. This reduces the number of recursive rule applications to its logarithm at the expense of introducing a logarithmic number of unfolded rules to the program. Efficiency crucially depends on the amount of simplification inside the unfolded rules. We prove a super-linear speedup theorem in the best case, i.e. speedup by more than a constant factor. Our optimization can lower the time complexity class of a program. In this paper, the super-linear speedup is within bounds: it holds up to an arbitrary but chosen upper bound on the number of recursive steps. We also report on the first results with a prototype implementation of repeated recursion unfolding. A simple program transformation completely removes recursion up to the chosen bound. The actual runtime improvement quickly reaches several orders of magnitude.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑